U.S. patent application number 16/185718 was filed with the patent office on 2020-05-14 for method and system for automated account opening decisioning.
This patent application is currently assigned to Bottomline Technologies (DE), Inc.. The applicant listed for this patent is Bottomline Technologies (DE), Inc.. Invention is credited to Peter Cousins, Leonardo Gil, Alexey Skosyrskiy.
Application Number | 20200151812 16/185718 |
Document ID | / |
Family ID | 70550698 |
Filed Date | 2020-05-14 |
![](/patent/app/20200151812/US20200151812A1-20200514-D00000.png)
![](/patent/app/20200151812/US20200151812A1-20200514-D00001.png)
![](/patent/app/20200151812/US20200151812A1-20200514-D00002.png)
![](/patent/app/20200151812/US20200151812A1-20200514-D00003.png)
![](/patent/app/20200151812/US20200151812A1-20200514-D00004.png)
United States Patent
Application |
20200151812 |
Kind Code |
A1 |
Gil; Leonardo ; et
al. |
May 14, 2020 |
METHOD AND SYSTEM FOR AUTOMATED ACCOUNT OPENING DECISIONING
Abstract
A method for using machine learning techniques to analyze past
decisions (made be administrators concerning account opening
requests) and to recommend whether an account opening request
should be allowed or denied.
Inventors: |
Gil; Leonardo; (Manchester,
NH) ; Cousins; Peter; (Rye, NH) ; Skosyrskiy;
Alexey; (Barrington, RI) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Bottomline Technologies (DE), Inc. |
Portsmouth |
NH |
US |
|
|
Assignee: |
Bottomline Technologies (DE),
Inc.
Portsmouth
NH
|
Family ID: |
70550698 |
Appl. No.: |
16/185718 |
Filed: |
November 9, 2018 |
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G06Q 40/025 20130101;
G06N 20/00 20190101 |
International
Class: |
G06Q 40/02 20060101
G06Q040/02; G06N 99/00 20060101 G06N099/00 |
Claims
1. A device for providing a recommendation concerning an account
opening request being reviewed, the computing device including:
memory comprising a non-transitory computer readable medium,
wherein: the memory stores past decisions made regarding past
account opening requests as past account opening records; the past
account opening records each include: a result comprising grant or
denial of the past account opening request associated with the
record; and properties of the past account opening request
associated with the record including a risk score determined for
the request associated with the record; the received account
opening request includes properties including a risk scored
determined for the received account opening request; and the memory
also stores a grant machine learning algorithm and a denial machine
learning algorithm; and circuitry configured to: access the past
account opening records stored in the memory; receive the account
opening request; determine a recommendation for granting or denying
the received account opening request, wherein the determination
comprises performing the following rules using the circuitry: rule
1: determine past grant records comprising the stored past account
opening records including a result of grant; rule 2: determine past
denial records comprising the stored past account opening records
including a result of denial; rule 3: configure the grant machine
learning algorithm, such that the grant machine learning algorithm
outputs a likelihood that an inputted account opening request is
granted; rule 4: configure the denial machine learning algorithm,
such that the denial machine learning algorithm outputs a
likelihood that an inputted account opening request is denied; rule
5: train the grant machine learning algorithm using the determined
past grant records, such that the outputted likelihood that an
inputted account opening request is granted depends on: the
properties of the inputted account opening request; and the results
and properties of the determined past grant records; rule 6: store
the trained grant machine learning algorithm in the memory; rule 7:
train the denial machine learning algorithm using the determined
past denial records, such that the outputted likelihood that an
inputted account opening request is denied depends on: the
properties of the inputted account opening request; and the results
and properties of the determined past denial records; rule 8: store
the trained denial machine learning algorithm in the memory; rule
9: input the received account opening request to the trained grant
machine learning algorithm and receive the likelihood of grant
output by the grant machine learning algorithm; rule 10: input the
received account opening request to the trained denial machine
learning algorithm and receive the likelihood of denial output by
the denial machine learning algorithm; and rule 11: calculate the
recommendation for granting the received account opening request,
denying the received account opening request, or no recommendation
based on the received likelihood of grant and the received
likelihood of denial; and output the recommendation for granting or
denying the received account opening request.
2. The device of claim 1, further comprising a display device,
wherein the circuitry is further configured to display on the
display device the outputted recommendation for granting or denying
the received account opening request.
3. The device of claim 2, wherein the circuitry is further
configured to cause the display device to display along with the
outputted recommendation at least one of the properties of the
received account opening request.
4. The device of claim 3, further comprising an input device for
receiving an input from a user of the device, wherein: the
circuitry is further configured to cause the display to display a
user interface along with the outputted recommendation and the at
least one of the properties of the received account opening
request; the user interface includes an input for selecting using
the input device a denial or a grant of the received account
opening request.
5. The device of claim 4, wherein the circuitry is additionally
configured to receive the selected input and identify the received
account opening request as denied or granted in accordance with the
received input.
6. The device of claim 1, wherein the recommendation for granting
or denying the received account opening request comprises a grant
score based on the received likelihood of grant and a deny score
based on the received likelihood of denial.
7. The device of claim 1, wherein the recommendation for granting
or denying the received account opening request comprises a total
score based on a combination of the received likelihood of grant
and the received likelihood of denial.
8. The device of claim 1, wherein: when the received likelihood of
grant is above a predetermined grant threshold, the outputting of
the recommendation for granting or denying the received account
opening request comprises identifying the received account opening
request as granted; and when the received likelihood of denial is
above a predetermined denial threshold, the outputting of the
recommendation for granting or denying the received account opening
request comprises identifying the received account opening request
as denied.
9. The device of claim 1, wherein: when the received likelihood of
grant is above a predetermined grant high threshold and the
received likelihood of denial is below a predetermined denial low
threshold, the outputting of the recommendation for granting or
denying the received account opening request comprises identifying
the received account opening request as granted; and when the
received likelihood of denial is above a predetermined denial high
threshold and the received likelihood of denial is below a
predetermined denial low threshold, the outputting of the
recommendation for granting or denying the received account opening
request comprises identifying the received account opening request
as denied.
10. The device of claim 1, wherein at least one of the grant
machine learning algorithm or the denial machine learning algorithm
comprises at least one of a neural network, a support vector
machine.
11. The device of claim 1, wherein the risk score is received from
a system configured to output a risk of fraud based on data
included in a received account opening request.
12. The device of claim 1, wherein the account opening request
being reviewed includes missing data, inaccurate data, or an
inconclusive risk score.
13. The device of claim 1, wherein the account opening request
comprises at least one of a request to open an account at a
financial institution or a request to add a service to an
account.
14. The device of claim 1, wherein the properties of the past
account opening record and the properties of the received account
opening request include at least one of a credit score, credit
history, an annual income, occupation, debit tools, history of
non-payment of accounts, past bankruptcy, investment portfolio,
savings amount, or investment amount.
15. A method for providing a recommendation concerning an account
opening request being reviewed using machine learning performed on
circuitry, the method comprising: using the circuitry, accessing
past decisions made regarding past account opening requests stored
as past account opening records in a memory comprising a
non-transitory computer readable medium, wherein: the past account
opening records each include: a result comprising grant or denial
of the past account opening request associated with the record; and
properties of the past account opening request associated with the
record including a risk score determined for the request associated
with the record; and the received account opening request includes
properties including a risk scored determined for the received
account opening request; receiving with the circuitry the account
opening request; using the circuitry, determining a recommendation
for granting or denying the received account opening request,
wherein the determination comprises performing the following rules
using the circuitry: rule 1: determine past grant records
comprising the stored past account opening records including a
result of grant; rule 2: determine past denial records comprising
the stored past account opening records including a result of
denial; rule 3: configure a grant machine learning algorithm stored
in the memory, such that the grant machine learning algorithm
outputs a likelihood that an inputted account opening request is
granted; rule 4: configure a denial machine learning algorithm
stored in the memory, such that the denial machine learning
algorithm outputs a likelihood that an inputted account opening
request is denied; rule 5: train the grant machine learning
algorithm using the determined past grant records, such that the
outputted likelihood that an inputted account opening request is
granted depends on: the properties of the inputted account opening
request; and the results and properties of the determined past
grant records; rule 6: store the trained grant machine learning
algorithm in the memory; rule 7: train the denial machine learning
algorithm using the determined past denial records, such that the
outputted likelihood that an inputted account opening request is
denied depends on: the properties of the inputted account opening
request; and the results and properties of the determined past
denial records; rule 8: store the trained denial machine learning
algorithm in the memory; rule 9: input the received account opening
request to the trained grant machine learning algorithm and receive
the likelihood of grant output by the grant machine learning
algorithm; rule 10: input the received account opening request to
the trained denial machine learning algorithm and receive the
likelihood of denial output by the denial machine learning
algorithm; and rule 11: calculate the recommendation for granting
or denying the received account opening request based on the
received likelihood of grant and the received likelihood of denial;
and using the circuitry, outputting the recommendation for granting
or denying the received account opening request.
16. The method of claim 15, further comprising displaying on a
display device the outputted recommendation for granting or denying
the received account opening request.
17. The method of claim 16, wherein the outputted recommendation is
displayed along with at least one of the properties of the received
account opening request.
18. The method of claim 17, further comprising: receiving from an
input device an input from a user; displaying a user interface
along with the outputted recommendation and the at least one of the
properties of the received account opening request, wherein the
user interface includes an input for selecting using the input
device a denial or a grant of the received account opening request;
and receiving the selected input and identifying using the
circuitry the received account opening request as denied or granted
in accordance with the received input.
19. The method of claim 1, wherein: when the received likelihood of
grant is above a predetermined grant threshold, the outputting of
the recommendation for granting or denying the received account
opening request comprises identifying the received account opening
request as granted; and when the received likelihood of denial is
above a predetermined denial threshold, the outputting of the
recommendation for granting or denying the received account opening
request comprises identifying the received account opening request
as denied.
20. The method of claim 1, wherein: when the received likelihood of
grant is above a predetermined grant high threshold and the
received likelihood of denial is below a predetermined denial low
threshold, the outputting of the recommendation for granting or
denying the received account opening request comprises identifying
the received account opening request as granted; and when the
received likelihood of denial is above a predetermined denial high
threshold and the received likelihood of denial is below a
predetermined grant low threshold, the outputting of the
recommendation for granting or denying the received account opening
request comprises identifying the received account opening request
as denied.
Description
TECHNICAL FIELD The present disclosure relates generally to machine
learning techniques and, more particularly, to using machine
learning algorithm to analyze past decisions.
BACKGROUND
[0001] When opening a new account or adding services to an account,
accounts often end up in a manual review process. A manual review
process may be initiated when a user fails to answer questions
accurately or completely when setting up an account or when a risk
provider identifies the account opener as risky (e.g., potentially
fraudulent). In manual review, human administrators review the
available information to determine if the account opening request
should be allowed or rejected.
SUMMARY
[0002] The present disclosures provides a method for using machine
learning techniques to analyze past decisions (made by
administrators concerning account opening requests) and to
recommend whether an account opening request should be allowed or
denied.
[0003] Currently, risk analyzers (e.g., fraud detection companies)
assess the risk of a new application based on, e.g., credit
history, sanctions lists, etc., but do not take into consideration
past decisions made by an organization. The present disclosure
analyzes past decisions to make new application decisions more
consistent (e.g., by a single administrator, across an
organization, etc.).
[0004] According to one aspect, there is provided a device for
providing a recommendation concerning an account opening request
being reviewed. The computing device includes memory and circuitry.
The memory includes a non-transitory computer readable medium and
stores past decisions made regarding past account opening requests
as past account opening records. The past account opening records
each include: a result comprising grant or denial of the past
account opening request associated with the record; and properties
of the past account opening request associated with the record
including a risk score determined for the request associated with
the record. The received account opening request includes
properties including a risk scored determined for the received
account opening request. The memory also stores a grant machine
learning algorithm and a denial machine learning algorithm. The
circuitry is configured to access the past account opening records
stored in the memory, receive the account opening request, and
determine a recommendation for granting or denying the received
account opening request. The determination comprises performing the
following rules using the circuitry. Rule 1: determine past grant
records comprising the stored past account opening records
including a result of grant. Rule 2: determine past denial records
comprising the stored past account opening records including a
result of denial. Rule 3: configure the grant machine learning
algorithm, such that the grant machine learning algorithm outputs a
likelihood that an inputted account opening request is granted.
Rule 4: configure the denial machine learning algorithm, such that
the denial machine learning algorithm outputs a likelihood that an
inputted account opening request is denied. Rule 5: train the grant
machine learning algorithm using the determined past grant records,
such that the outputted likelihood that an inputted account opening
request is granted depends on: the properties of the inputted
account opening request and the results and properties of the
determined past grant records. Rule 6: store the trained grant
machine learning algorithm in the memory. Rule 7: train the denial
machine learning algorithm using the determined past denial
records, such that the outputted likelihood that an inputted
account opening request is denied depends on: the properties of the
inputted account opening request; and the results and properties of
the determined past denial records. Rule 8: store the trained
denial machine learning algorithm in the memory. Rule 9: input the
received account opening request to the trained grant machine
learning algorithm and receive the likelihood of grant output by
the grant machine learning algorithm. Rule 10: input the received
account opening request to the trained denial machine learning
algorithm and receive the likelihood of denial output by the denial
machine learning algorithm. Rule 11: calculate the recommendation
for granting the received account opening request, denying the
received account opening request, or no recommendation based on the
received likelihood of grant and the received likelihood of denial.
The circuitry is also configured to output the recommendation for
granting or denying the received account opening request.
[0005] Alternatively or additionally, further comprising a display
device. The circuitry is further configured to display on the
display device the outputted recommendation for granting or denying
the received account opening request.
[0006] Alternatively or additionally, the circuitry is further
configured to cause the display device to display along with the
outputted recommendation at least one of the properties of the
received account opening request.
[0007] Alternatively or additionally, further comprising an input
device for receiving an input from a user of the device. The
circuitry is further configured to cause the display to display a
user interface along with the outputted recommendation and the at
least one of the properties of the received account opening
request. The user interface includes an input for selecting using
the input device a denial or a grant of the received account
opening request.
[0008] Alternatively or additionally, the circuitry is additionally
configured to receive the selected input and identify the received
account opening request as denied or granted in accordance with the
received input.
[0009] Alternatively or additionally, the recommendation for
granting or denying the received account opening request comprises
a grant score based on the received likelihood of grant and a deny
score based on the received likelihood of denial.
[0010] Alternatively or additionally, the recommendation for
granting or denying the received account opening request comprises
a total score based on a combination of the received likelihood of
grant and the received likelihood of denial.
[0011] Alternatively or additionally, when the received likelihood
of grant is above a predetermined grant threshold, the outputting
of the recommendation for granting or denying the received account
opening request comprises identifying the received account opening
request as granted. When the received likelihood of denial is above
a predetermined denial threshold, the outputting of the
recommendation for granting or denying the received account opening
request comprises identifying the received account opening request
as denied.
[0012] Alternatively or additionally, when the received likelihood
of grant is above a predetermined grant high threshold and the
received likelihood of denial is below a predetermined denial low
threshold, the outputting of the recommendation for granting or
denying the received account opening request comprises identifying
the received account opening request as granted. When the received
likelihood of denial is above a predetermined denial high threshold
and the received likelihood of denial is below a predetermined
denial low threshold, the outputting of the recommendation for
granting or denying the received account opening request comprises
identifying the received account opening request as denied.
[0013] Alternatively or additionally, at least one of the grant
machine learning algorithm or the denial machine learning algorithm
comprises at least one of a neural network, a support vector
machine.
[0014] Alternatively or additionally, the risk score is received
from a system configured to output a risk of fraud based on data
included in a received account opening request.
[0015] Alternatively or additionally, the account opening request
being reviewed includes missing data, inaccurate data, or an
inconclusive risk score.
[0016] Alternatively or additionally, the account opening request
comprises at least one of a request to open an account at a
financial institution or a request to add a service to an
account.
[0017] Alternatively or additionally, the properties of the past
account opening record and the properties of the received account
opening request include at least one of a credit score, credit
history, an annual income, occupation, debit tools, history of
non-payment of accounts, past bankruptcy, investment portfolio,
savings amount, or investment amount.
[0018] The present disclosure also provides a method for providing
a recommendation concerning an account opening request being
reviewed using machine learning. The method includes (using
circuitry) accessing past decisions made regarding past account
opening requests stored as past account opening records in a memory
comprising a non-transitory computer readable medium. The past
account opening records each include: a result comprising grant or
denial of the past account opening request associated with the
record; and properties of the past account opening request
associated with the record including a risk score determined for
the request associated with the record. The received account
opening request includes properties including a risk scored
determined for the received account opening request. The method
also includes receiving with the circuitry the account opening
request. The method further includes (using the circuitry)
determining a recommendation for granting or denying the received
account opening request. The determination comprises performing the
following rules using the circuitry. Rule 1: determine past grant
records comprising the stored past account opening records
including a result of grant. Rule 2: determine past denial records
comprising the stored past account opening records including a
result of denial. Rule 3: configure a grant machine learning
algorithm stored in the memory, such that the grant machine
learning algorithm outputs a likelihood that an inputted account
opening request is granted. Rule 4: configure a denial machine
learning algorithm stored in the memory, such that the denial
machine learning algorithm outputs a likelihood that an inputted
account opening request is denied. Rule 5: train the grant machine
learning algorithm using the determined past grant records, such
that the outputted likelihood that an inputted account opening
request is granted depends on: the properties of the inputted
account opening request; and the results and properties of the
determined past grant records. Rule 6: store the trained grant
machine learning algorithm in the memory. Rule 7: train the denial
machine learning algorithm using the determined past denial
records, such that the outputted likelihood that an inputted
account opening request is denied depends on: the properties of the
inputted account opening request; and the results and properties of
the determined past denial records. Rule 8: store the trained
denial machine learning algorithm in the memory. Rule 9: input the
received account opening request to the trained grant machine
learning algorithm and receive the likelihood of grant output by
the grant machine learning algorithm. Rule 10: input the received
account opening request to the trained denial machine learning
algorithm and receive the likelihood of denial output by the denial
machine learning algorithm. Rule 11: calculate the recommendation
for granting or denying the received account opening request based
on the received likelihood of grant and the received likelihood of
denial. The method also includes (using the circuitry) outputting
the recommendation for granting or denying the received account
opening request.
[0019] Alternatively or additionally, further comprising displaying
on a display device the outputted recommendation for granting or
denying the received account opening request.
[0020] Alternatively or additionally, the outputted recommendation
is displayed along with at least one of the properties of the
received account opening request.
[0021] Alternatively or additionally, the method further comprising
receiving from an input device an input from a user and displaying
a user interface along with the outputted recommendation and the at
least one of the properties of the received account opening
request. The user interface includes an input for selecting using
the input device a denial or a grant of the received account
opening request. The method also includes receiving the selected
input and identifying using the circuitry the received account
opening request as denied or granted in accordance with the
received input.
[0022] Alternatively or additionally, when the received likelihood
of grant is above a predetermined grant threshold, the outputting
of the recommendation for granting or denying the received account
opening request comprises identifying the received account opening
request as granted. When the received likelihood of denial is above
a predetermined denial threshold, the outputting of the
recommendation for granting or denying the received account opening
request comprises identifying the received account opening request
as denied.
[0023] Alternatively or additionally, when the received likelihood
of grant is above a predetermined grant high threshold and the
received likelihood of denial is below a predetermined denial low
threshold, the outputting of the recommendation for granting or
denying the received account opening request comprises identifying
the received account opening request as granted. When the received
likelihood of denial is above a predetermined denial high threshold
and the received likelihood of denial is below a predetermined
grant low threshold, the outputting of the recommendation for
granting or denying the received account opening request comprises
identifying the received account opening request as denied.
[0024] While a number of features are described herein with respect
to embodiments of the invention; features described with respect to
a given embodiment also may be employed in connection with other
embodiments. The following description and the annexed drawings set
forth certain illustrative embodiments of the invention. These
embodiments are indicative, however, of but a few of the various
ways in which the principles of the invention may be employed.
Other objects, advantages and novel features according to aspects
of the invention will become apparent from the following detailed
description when considered in conjunction with the drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
[0025] The annexed drawings, which are not necessarily to scale,
show various aspects of the invention in which similar reference
numerals are used to indicate the same or similar parts in the
various views.
[0026] FIG. 1 is a schematic diagram of an exemplary computing
device according to the invention.
[0027] FIG. 2 is a ladder diagram depicting movement of information
between the circuitry and memory of FIG. 1.
[0028] FIG. 3 is a block diagram depicting training of the grant
machine learning algorithm and the denial machine learning
algorithm.
[0029] FIG. 4 is a block diagram depicting classification of an
account opening request by the grant machine learning algorithm and
the denial machine learning algorithm.
[0030] FIG. 5 is a flow diagram of a method for providing a
recommendation concerning an account opening request according to
the invention.
DETAILED DESCRIPTION
[0031] The present invention is now described in detail with
reference to the drawings. In the drawings, each element with a
reference number is similar to other elements with the same
reference number independent of any letter designation following
the reference number. In the text, a reference number with a
specific letter designation following the reference number refers
to the specific element with the number and letter designation and
a reference number without a specific letter designation refers to
all elements with the same reference number independent of any
letter designation following the reference number in the
drawings.
[0032] The present invention provides a device including circuitry
and memory. The circuitry uses machine learning techniques to
analyze past decisions concerning account opening requests that are
stored in the memory. The circuitry outputs a recommendation
regarding whether an account opening request should be allowed or
denied.
[0033] Turning to FIG. 1, an exemplary computing device 10 for
providing a recommendation concerning an account opening request
being reviewed is shown. The device 10 includes memory 12 and
circuitry 14. The device 10 may also include a communication
interface 16, a display 18, and/or an input 20. The memory 12
stores past decisions made regarding past account opening requests
as past account opening records 26. The circuitry 14 accesses the
past account opening records 26 and determines a recommendation 29
for granting or denying a received account opening request based
upon the past account opening records 26.
[0034] The account opening request 27 may be received by the
circuitry from the communication interface 16. The account opening
request 27 may include at least one of a request to open an
existing account at a financial institution or a request to add a
service to an account. The account opening request 27 may be passed
to the computing device 10 for making a recommendation 29 when the
account opening request 27 includes missing data, inaccurate data,
or an inconclusive risk score is received, e.g., from a risk score
system 40.
[0035] As described above, the memory 12 stores a grant machine
learning algorithm 28, a denial machine learning algorithm 30, and
past decisions made regarding past account opening requests as past
account opening records 26. The past account opening records 26
each include a result 32 and properties 34 of the past account
opening request associated with the record 26. The result 32
includes a grant or denial of the past account opening request
associated with the record 26. The properties 34 of the past
account opening request associated with the record include a risk
score determined for the request associated with the record 26.
[0036] The past account opening records 26 may be limited to
records associated with a particular administrator that has been
assigned to review (i.e., is associated with) the received account
opening request 27. In this way, decisions made by the particular
administrator may only be used when making a recommendation 29 and
the decisions made by individual administrators may be improved.
Conversely, the past account opening records 26 may include records
associated with a particular organization (e.g., a specific
financial institution) without regard to the particular
administrator associated with the account opening request 27. In
this way, decisions made across the particular organization may be
made more consistent across different administrators.
[0037] The past account opening records 26 may also be limited in
time. For example, only recent past account opening records 26 may
be used (e.g., within the last year, 6 months, 1 month, etc.) to
ensure that recent changes in organization procedures are adapted
and used consistently. As another example, past account opening
records 26 over a longer duration of time (e.g., 2 years, 5 years,
10 years, all available records, etc.) may be used to improve
consistency of decision over time.
[0038] The properties 44 of the past account opening records 26 and
the properties 44 of the received account opening request 27 may
include at least one of a credit score, credit history, an annual
income, occupation, debit tools, history of non-payment of
accounts, past bankruptcy, investment portfolio, savings amount, or
investment amount. As will be understood by one of ordinary skill
in the art, the properties 34, 44 may include any suitable data for
making a determination regarding granting or denying an account
opening request.
[0039] The risk score may be received from a system 40 configured
to output a risk of fraud based on data (e.g., the properties 44)
included in a received account opening request 27. For example, the
system 40 could be an outside third party, an internal tool, or a
combination thereof.
[0040] As will be understood by one of ordinary skill in the art,
the memory 12 may comprise one or more of a buffer, a flash memory,
a hard drive, a removable media, a volatile memory, a non-volatile
memory, a random access memory (RAM), or other suitable device. In
a typical arrangement, the memory 12 may include a non-volatile
memory for long term data storage and a volatile memory that
functions as system memory for the circuitry 14. The memory 12 may
exchange data with the circuitry 14 over a data bus. Accompanying
control lines and an address bus between the memory 12 and the
circuitry 14 also may be present. The memory 12 may be considered a
non-transitory computer readable medium.
[0041] Turning to FIG. 2, the circuitry 14 is configured to access
the past account opening records 26 stored in the memory 12 and to
receive an account opening request 27. The circuitry 14 is also
configured to determine a recommendation 29 for granting or denying
the received account opening request and to output the
recommendation 29 for granting or denying the received account
opening request 27. The recommendation is determined by performing
the following rules using the circuitry 14.
[0042] In rule 1, the circuitry 14 determines past grant records 31
comprising the stored past account opening records 26 including a
result 32 of grant. That is, the circuitry 14 determines the stored
past account opening records 26 where the account opening record
was granted (e.g., an account was opened). In rule 2, the circuitry
determines past denial records 33 comprising the stored past
account opening records 26 including a result 32 of denial. That
is, the circuitry 14 determines the stored past account opening
records 26 where the account opening record was denied (e.g., an
account was not opened).
[0043] In rule 3, the circuitry 14 configures the grant machine
learning algorithm 28, such that the grant machine learning
algorithm 28 outputs a likelihood 35 that an inputted account
opening request 27 is granted. In rule 4, the circuitry 14
configures the denial machine learning algorithm 30, such that the
denial machine learning algorithm 28 outputs a likelihood 37 that
an inputted account opening request 27 is denied. The received
account opening request 27 includes properties including a risk
scored determined for the received account opening request 27.
[0044] Rules 3 and 4 may be performed by a separate electronic
device (e.g., a computer) at any time prior to training the machine
learning algorithms 28, 30 by the circuitry 14. For example, the
structure of the machine learning algorithms 28, 30 may already be
determined and the circuitry 14 may simply access the machine
learning algorithms 28, 30 to perform rules 3 and 4.
[0045] At least one of the grant machine learning algorithm 28 or
the denial machine learning algorithm 30 may be a neural network, a
support vector machine, or any other suitable machine learning
algorithm. The grant machine learning algorithm 28 or the denial
machine learning algorithm 30 may utilize supervised learning,
unsupervised learning, or semi-supervised learning.
[0046] The grant machine learning algorithm 28 and/or the denial
machine learning algorithm 30 may comprise a neural network. As an
example, the grant machine learning algorithm 28 and/or the denial
machine learning algorithm 30 may comprise a bidirectional
recurrent neural network (BRNN) and the machine learning
algorithm(s) 28, 30 may be configured to output a Boolean result.
For example, the grant machine learning algorithm 28 outputting a
"1" may indicate that an account opening request 29 should be
granted and the grant machine learning algorithm 28 outputting a
"0" may indicate that an account opening request 29 should not be
opened or "no opinion" regarding whether the account opening
request 27 should be granted. Similarly, the denial machine
learning algorithm 30 outputting a "1" may indicate that an account
opening request 29 should be denied and the denial machine learning
algorithm 30 outputting a "0" may indicate that an account opening
request 29 should not be denied or "no opinion" regarding whether
the account opening request 27 should be denied.
[0047] In another example, the machine learning algorithm(s) 28, 30
may output a value within a range (e.g., between 0 and 1). For
example, the closer the value outputted by the grant machine
learning algorithm 28 is to one end of the range (e.g., "1") the
more likely that an account opening request 29 should be granted.
Similarly, the closer the value outputted by the denial machine
learning algorithm 28 is to one end of the range (e.g., "1") the
more likely that an account opening request 29 should be
denied.
[0048] Turning to FIGS. 2-4, in rule 5, the circuitry 14 trains the
grant machine learning algorithm 28 using the determined past grant
records 31, such that the outputted likelihood 35 that an inputted
account opening request is granted depends on: (1) the properties
44 of the inputted account opening request 27 and (2) the results
32 and properties 34 of the determined past grant records 31. In
rule 6, the circuitry stores the trained grant machine learning
algorithm 28 in the memory 12.
[0049] In the drawings reference numerals 28a and 28b are used to
differentiate between the untrained and trained grant machine
learning algorithm, respectively, when both the trained and
untrained grant machine learning algorithm are shown in the same
figure. In figures where both the trained and untrained grant
machine learning algorithms are not both present, only reference
numeral 28 may beused. Similar comments apply regarding the denial
machine learning algorithm and reference numerals 30a and 30b.
[0050] In rule 7, the circuitry 14 trains the denial machine
learning algorithm 30 using the determined past denial records 33,
such that the outputted likelihood that an inputted account opening
request is denied 37 depends on: (1) the properties 44 of the
inputted account opening request 27 and (2) the results 32 and
properties 34 of the determined past denial records 33. In rule 8,
the circuitry 14 causes the trained denial machine learning
algorithm 28b to be stored in the memory 12.
[0051] Turning to FIGS. 3 and 4, rules 1-8 may be performed by a
separate device 10a from the device 10b performing rules 9 and 10.
For example, the computing device 10 may comprise two separate
devices 10a, 10b. Each device 10a, 10b may include circuitry 14a,
14b. The first device 10a of the two separate devices 10a, 10b may
perform rules 1-8 using circuitry 14a local to the first device
10a. The circuitry 14a may store the trained machine learning
algorithms 18, 30 in memory 14 accessible (e.g., accessible over a
network) by the circuitry 14a of the first device 10 and the
circuitry 14b of the second device 10b of the two separate devices
10a, 10b. The circuitry 14b of the second device 10b may then
access the trained machine learning algorithms 18, 30 in the memory
14 and input the account opening request 27 in rules 9 and 10 to
determine the likelihood of grant 35 and the likelihood of denial
37. Alternatively, rules 1-10 may all be performed by circuitry 14
located in the same device 10.
[0052] In rule 9, the circuitry 14 inputs the received account
opening request 27 to the trained grant machine learning algorithm
28 and receives the likelihood of grant 35 output from the grant
machine learning algorithm 28. Similarly, in rule 10, the circuitry
14 inputs the received account opening request 27 to the trained
denial machine learning algorithm 30 and receives the likelihood of
denial 37 output by the denial machine learning algorithm 30.
[0053] While receiving the account opening request 27 is described
prior to rules 1-8, the account opening request 27 may be received
by the circuitry 14 after performance of rules 1-8.
[0054] In rule 11, the circuitry 14 calculates the recommendation
29 for (1) granting the received account opening request 27, (2)
denying the received account opening request 27, or (3) no
recommendation. The recommendation 29 is calculated based on the
received likelihood of grant 35 and the received likelihood of
denial 37.
[0055] As will be understood by one of ordinary skill in the art,
the circuitry 14 may have various implementations. For example, the
circuitry 14 may include any suitable device, such as a processor
(e.g., CPU), programmable circuit, integrated circuit, memory and
I/O circuits, an application specific integrated circuit,
microcontroller, complex programmable logic device, other
programmable circuits, or the like. The circuitry 14 may also
include a non-transitory computer readable medium, such as random
access memory (RAM), a read-only memory (ROM), an erasable
programmable read-only memory (EPROM or Flash memory), or any other
suitable medium. Instructions for performing the method 100
described below may be stored in the non-transitory computer
readable medium and executed by the circuitry 14. The circuitry 14
may be communicatively coupled to the memory 12 and a communication
interface 16 through a system bus, mother board, or using any other
suitable structure known in the art.
[0056] Turning back to FIG. 1, the device 10 may also include a
display device 18. The circuitry 14 may be configured to cause the
display device 18 to display the outputted recommendation 29 for
granting or denying the received account opening request 27. For
example, the circuitry 14 may be further configured to cause the
display device 18 to display along with the outputted
recommendation 29 at least one of the properties 44 of the received
account opening request 27. In this way, an administrator reviewing
account opening request 27 may view the properties 44 needed to
make an informed decision regarding granting or denying the request
along with the recommendation 29.
[0057] As will be understood by one of ordinary skill in the art,
the display device 18 may have various implementations. For
example, the display device 18 may comprise any suitable device for
displaying information, such as a liquid crystal display, light
emitting diode display, a CRT display, an organic light emitting
diode (OLED) display, a computer monitor, a television, a phone
screen, or the like. The display device 18 may also include an
interface (e.g., HDMI input, USB input, etc.) for receiving
information to be displayed.
[0058] The device 10 may also include an input device 20 for
receiving an input from a user of the device 10. For example, when
displaying the recommendation 29 and the at least one of the
properties 44 of the received account opening request 27, the user
interface 36 may include an input for selecting (using the input
device 20) a denial or a grant of the received account opening
request. The circuitry 14 may be configured to receive the selected
input and identify the received account opening request as denied
or granted in accordance with the received input. The circuitry 14
may then cause the communication interface 16 to transmit the
selected input (i.e., to deny or grant the account opening request
27). The circuitry 14 may also cause the account opening request 27
to be stored as a past account opening record 26, including (as
properties 34) the properties 44 of the account opening request 27
and (as the result 32) the selected input.
[0059] As the past account opening records 26 are updated by the
circuitry 14 to include account opening requests 27 that have been
decided by an administrator (i.e., that have a result 32), the
training of the machine learning algorithms 28, 30 may be updated
(also referred to as the machine learning algorithms 28, 30 being
updated). In this way, performance of the machine learning
algorithms 28, 30 may be continuously or periodically updated. For
example, the machine learning algorithms 28, 30 may be updated
daily, weekly, monthly, or based on the number of new past account
opening records 26 (e.g., every 100, 250, or 1000 new past account
opening records 26).
[0060] As will be understood by one of ordinary skill in the art,
the input device 20 may have various implementations. For example,
the input device 20 may comprise any suitable device for inputting
data into an electronic device, such as a keyboard, mouse,
trackpad, touch screen (e.g., as part of the display device 18),
microphone, or the like.
[0061] As will be understood by one of ordinary skill in the art,
the communication interface 16 may comprise a wireless network
adaptor, an Ethernet network card, or any suitable device that
provides an interface between the device 10 and a network. The
communication interface 16 may be communicatively coupled to the
memory 12, such that the communication interface 16 is able to send
data stored on the memory 12 across the network and store received
data on the memory 12. The communication interface 16 may also be
communicatively coupled to the circuitry 14 such that the circuitry
14 is able to control operation of the communication interface 16.
The communication interface 16, memory 12, and circuitry 14 may be
communicatively coupled through a system bus, mother board, or
using any other suitable manner as will be understood by one of
ordinary skill in the art.
[0062] The recommendation 29 for granting or denying the received
account opening request 27 may comprise a grant score based on the
received likelihood of grant 35 and a deny score based on the
received likelihood of denial 37. For example, (instead of simply
being grant, deny, or no decision), the recommendation 29 for
granting or denying the received account opening request may
comprise a total score based on a combination of the received
likelihood of grant 35 and the received likelihood of denial
37.
[0063] The recommendation 29 may also be calculated from the
received likelihood of grant 35 and the received likelihood of
denial 37 using any suitable method. For example, the
recommendation 29 may be calculated using a machine learning
algorithm. In this example, a recommendation machine learning
algorithm may take as an input the received likelihood of grant 35
and the received likelihood of denial 37 and output the
recommendation 29. The recommendation machine learning algorithm
may be trained using the past account opening records 26 (including
the result 32) and the likelihood of grant and denial 35, 37 output
by the grant and denial machine learning algorithms 28, 30,
respectively. Because the output of the recommendation machine
learning algorithm depends on the output of the grant and denial
machine learning algorithms 28, 30, the grant and denial machine
learning algorithms 28, 30 may be trained before and separately
from the recommendation machine learning algorithm. Conversely, the
recommendation, grant, and denial machine learning algorithms may
all be trained at the same time.
[0064] As opposed to using a separate machine learning algorithm,
the recommendation 29 may be calculated from the likelihood of
grant 35 and the likelihood of denial 37 using rules. As an
example, when the received likelihood of grant 35 is above a
predetermined grant threshold, the outputted recommendation 29 may
be to identify the received account opening request 27 as granted.
Alternatively, when the received likelihood of denial 37 is above a
predetermined denial threshold, the outputted recommendation 29 may
be to identify the received account opening request 27 as denied.
If both the received likelihood of grant 35 and the received
likelihood of denial 37 are both above the predetermined grant
threshold and the predetermined denial threshold, respectively,
then an inconclusive recommendation 29 may be outputted. That is,
the device 10 may indicate that there is not a recommended to grant
or deny the received account opening request 27.
[0065] As another example, when the received likelihood of grant 35
is above a predetermined grant high threshold and the received
likelihood of denial is below a predetermined denial low threshold,
the outputted recommendation 29 may be to identify the received
account opening request 27 as granted. When the received likelihood
of denial 37 is above a predetermined denial high threshold and the
received likelihood of grant is below a predetermined grant low
threshold, the outputted recommendation 29 may be to identify the
received account opening request 27 as denied. If (1) the received
likelihood of grant 35 is between the predetermined grant high
threshold and the predetermined grant low threshold and (2) the
received likelihood of denial 37 is between the predetermined
denial high threshold and the predetermined denial low threshold,
then no (or an inconclusive) recommendation 29 may be
outputted.
[0066] The recommendation 29 may additionally include a confidence
measure dependent upon the likelihood of grant 35 and the
likelihood of denial 37. For example, (1) the grant machine
learning algorithm 28 may output a likelihood of grant 35 and a
grant confidence measure and (2) the denial machine learning
algorithm 30 may output a likelihood of denial 35 and a denial
confidence measure. The grant confidence measure and the denial
confidence measure may be an indication of how similar the received
account opening request 27 is to past account opening records 26.
For example, the properties 44 of the received account opening
request 27 may be compared to the properties 34 of the past account
opening records 26. The more similar the received account opening
request 27 is to one or more of the past account opening records
26, the higher the confidence score may be. For example, if the
received account opening request 27 is similar to one or more past
account opening records that were denied, then the denial
confidence score may be higher.
[0067] Alternatively or additionally, the device 10 may determine a
performance measure. The performance measure may be based upon the
number of times that an administrator agrees with the
recommendation 29. For example, if the recommendation 29 is to deny
and the administrator agrees with the recommendation and denies the
account opening request, then the performance measure may increase.
Conversely, if the recommendation is to deny and the administrator
disagrees with the recommendation and allows the account opening
request, then the performance measure may decrease.
[0068] The performance measure may be adjusted to measure lifetime
performance or performance during a particular time span. For
example, the performance measure may be based upon only the past
three months (or any suitable duration of time) and all previous
performance may be disregarded.
[0069] If the performance measure and/or the confidence measure is
above a given threshold, then the circuitry 14 may make the final
decision regarding allowing or denying the received account opening
request 27. That is, if the performance measure and/or the
confidence measure are above the given threshold to grant the
received account opening request 27, then the received account
opening request 27 may be granted by the circuitry 14 without
intervention by a human administrator.
[0070] Turning to FIG. 5, a method 100 for providing a
recommendation 29 concerning an account opening request 27 being
reviewed using machine learning performed on circuitry 14 is
shown.
[0071] In reference block 102, the circuitry 14 receives the
account opening request 27. In reference block 104, the circuitry
14 accesses past decisions made regarding past account opening
requests stored as past account opening records 26 in a memory 12
comprising a non-transitory computer readable medium. As described
above, the past account opening records 26 each include a result 32
and properties 34. The result comprises grant or denial of the past
account opening request associated with the record 26. The
properties 34 of the past account opening request associated with
the record 26 include a risk score determined for the request
associated with the record 26. The received account opening request
27 also includes properties 44 including a risk scored determined
for the received account opening request 27.
[0072] In reference blocks 106-128, the circuitry 14 determines a
recommendation for granting or denying the received account opening
request 27.
[0073] In reference block 106, past grant records 31 comprising the
stored past account opening records including a result of grant are
determined. Similarly, in reference block 108, past denial records
33 comprising the stored past account opening records including a
result of denial are determined.
[0074] In reference block 110, a grant machine learning algorithm
28 stored in the memory 12 is configured, such that the grant
machine learning algorithm 28 outputs a likelihood that an inputted
account opening request is granted 35. Similarly, in reference
block 112, a denial machine learning algorithm 30 stored in the
memory 12 is configured, such that the denial machine learning
algorithm 30 outputs a likelihood that an inputted account opening
request is denied 37. As described above, reference blocks 110 and
112 may be performed separately at any point in time prior to
performing reference blocks 114 and 116.
[0075] In reference block 114, the grant machine learning algorithm
28 is trained using the determined past grant records, such that
the outputted likelihood 35 that an inputted account opening
request is granted depends on (1) the properties of the inputted
account opening request 27 and (2) the results 32 and properties 34
of the determined past grant records 26. In reference block 116,
the trained grant machine learning algorithm 28 is stored in the
memory 12.
[0076] In reference block 115, the denial machine learning
algorithm 30 is trained using the determined past denial records,
such that the outputted likelihood 37 that an inputted account
opening request is denied depends on: (1) the properties of the
inputted account opening request and (2) the results 32 and
properties 34 of the determined past denial records. In reference
block 118, the trained denial machine learning algorithm 30 is
stored in the memory 12.
[0077] In reference block 120, the received account opening request
27 is inputted into the trained grant machine learning algorithm
28. Similarly in reference block 122, the received account opening
request 27 is inputted into the trained denial machine learning
algorithm 30.
[0078] In reference block 124, the likelihood of grant is outputted
by the grant machine learning algorithm 28. Similarly, in reference
block 126, the likelihood of denial is output by the denial machine
learning algorithm 30.
[0079] In reference block 128, the recommendation 29 for granting
or denying the received account opening request 27 is calculated
based on the received likelihood of grant 26 and the received
likelihood of denial 37. In reference block 130, the circuitry
outputs the recommendation 29 for granting or denying the received
account opening request. As described above, outputting the
recommendation 29 may comprise displaying the recommendation 29,
storing the recommendation 29, or transmitting (e.g., via the
communication interface 16 over a network) the recommendation
29.
[0080] While reference block 102 is shown as occurring before
reference blocks 104-118, reference block 102 may occur before
during or after any of reference block 104-118. For example, the
machine learning algorithms 28, 30 may be trained and stored in the
memory 12 prior to receiving the account opening request 27.
[0081] It should be appreciated that many of the elements discussed
in this specification may be implemented in a hardware circuit(s),
a processor executing software code or instructions which are
encoded within computer readable media accessible to the processor,
or a combination of a hardware circuit(s) and a processor or
control block of an integrated circuit executing machine readable
code encoded within a computer readable media. As such, the term
circuit, module, server, application, or other equivalent
description of an element as used throughout this specification is,
unless otherwise indicated, intended to encompass a hardware
circuit (whether discrete elements or an integrated circuit block),
a processor or control block executing code encoded in a computer
readable media, or a combination of a hardware circuit(s) and a
processor and/or control block executing such code.
[0082] All ranges and ratio limits disclosed in the specification
and claims may be combined in any manner. Unless specifically
stated otherwise, references to "a," "an," and/or "the" may include
one or more than one, and that reference to an item in the singular
may also include the item in the plural.
[0083] Although the invention has been shown and described with
respect to a certain embodiment or embodiments, equivalent
alterations and modifications will occur to others skilled in the
art upon the reading and understanding of this specification and
the annexed drawings. In particular regard to the various functions
performed by the above described elements (components, assemblies,
devices, compositions, etc.), the terms (including a reference to a
"means") used to describe such elements are intended to correspond,
unless otherwise indicated, to any element which performs the
specified function of the described element (i.e., that is
functionally equivalent), even though not structurally equivalent
to the disclosed structure which performs the function in the
herein illustrated exemplary embodiment or embodiments of the
invention. In addition, while a particular feature of the invention
may have been described above with respect to only one or more of
several illustrated embodiments, such feature may be combined with
one or more other features of the other embodiments, as may be
desired and advantageous for any given or particular
application.
* * * * *